Fix PEFT and update README

README.md

@@ -1,113 +1,122 @@
 ---
-title: Kora API
-emoji: 🚀
-colorFrom: blue
-colorTo: purple
+license: mit
+title: KORA AI
+emoji: 👀
+colorTo: blue
+short_description: KORA is the intelligent AI assistant for the Kontyra ecosyst
 sdk: docker
-pinned: false
 ---
-# KORA API Backend (FastAPI + Transformers)
 
-Production-ready AI backend scaffold for KORA, optimized for Docker and Hugging Face Docker Spaces.
+# KORA AI
+
+KORA is the intelligent AI assistant for the Kontyra ecosystem.
+
+This Space hosts the KORA application and related AI services.
+
+
+## Overview
+
+This repository contains the backend infrastructure for KORA, built with FastAPI, Transformers, and Docker. The API is designed to be compatible with OpenAI-style chat completion endpoints while supporting custom system prompts, knowledge-base integration, and streaming responses.
 
 ## Features
 
-- OpenAI-compatible endpoint: `POST /v1/chat/completions`
-- Streaming token responses using SSE
-- CPU-safe Transformers + PyTorch runtime with `microsoft/Phi-3-mini-4k-instruct`
-- Automatic KORA system-prompt injection on every request
-- KONTYRA knowledge-base context injection
-- Async architecture with centralized logging
-- Environment-driven configuration
+* OpenAI-compatible API endpoint
+* FastAPI-based architecture
+* Streaming responses via Server-Sent Events (SSE)
+* Docker-ready deployment
+* Environment-based configuration
+* Custom KORA system prompt injection
+* Knowledge-base context integration
+* Scalable backend foundation for future Kontyra products
+
+## Technology Stack
+
+* Python
+* FastAPI
+* Transformers
+* PyTorch
+* Docker
+* Hugging Face
+
+## API Endpoint
+
+```http
+POST /v1/chat/completions
+```
 
-## Folder Structure
+### Example Request
+
+```json
+{
+  "model": "KORA",
+  "stream": true,
+  "messages": [
+    {
+      "role": "user",
+      "content": "Who created KORA?"
+    }
+  ]
+}
+```
+
+## Project Structure
 
 ```text
-apps/api/
-├── app/
-│   ├── routes/
-│   │   └── chat.py
-│   ├── services/
-│   │   ├── model_service.py
-│   │   └── prompt_service.py
-│   └── utils/
-│       └── config.py
+app/
+├── routes/
+├── services/
+├── utils/
 ├── prompts/
-│   └── system_prompt.txt
 ├── knowledge/
-│   └── kontyra.md
-├── Dockerfile
 ├── main.py
-└── requirements.txt
+├── requirements.txt
+└── Dockerfile
 ```
 
 ## Environment Variables
 
-Create `.env` in `apps/api/` (optional):
-
 ```bash
-APP_NAME="KORA AI Backend"
+APP_NAME="KORA AI"
 APP_ENV="production"
 LOG_LEVEL="INFO"
+
 MODEL_NAME="microsoft/Phi-3-mini-4k-instruct"
-TRUST_REMOTE_CODE=true
-LOW_CPU_MEM_USAGE=true
-MAX_INPUT_TOKENS=3072
-TORCH_NUM_THREADS=0
-TORCH_NUM_INTEROP_THREADS=0
+
 DEFAULT_TEMPERATURE=0.7
 DEFAULT_TOP_P=0.9
 DEFAULT_MAX_TOKENS=512
-```
 
-## Local Setup
+MAX_INPUT_TOKENS=3072
 
-```bash
-cd apps/api
-python -m venv .venv
-source .venv/bin/activate
-pip install -r requirements.txt
-uvicorn main:app --host 0.0.0.0 --port 8000
+TRUST_REMOTE_CODE=true
+LOW_CPU_MEM_USAGE=true
 ```
 
-## Docker Build + Run
-
-```bash
-cd apps/api
-docker build -t kora-api .
-docker run --rm -p 7860:7860 \
-  -e PORT=7860 \
-  -e MODEL_NAME=microsoft/Phi-3-mini-4k-instruct \
-  kora-api
-```
+## Deployment
 
-## Hugging Face Docker Spaces Startup
+This Space is configured as a Docker Space and automatically starts the KORA API service when deployed.
 
-Use the included `Dockerfile`. Hugging Face sets `PORT`, and the container starts with:
+The application listens on the port provided by Hugging Face:
 
 ```bash
 uvicorn main:app --host 0.0.0.0 --port ${PORT:-7860}
 ```
 
-## API Example
+## Vision
 
-### Request
+KORA serves as the intelligence layer of the Kontyra ecosystem and is being developed to support future products including:
 
-```json
-{
-  "model": "microsoft/Phi-3-mini-4k-instruct",
-  "stream": true,
-  "messages": [
-    {"role": "user", "content": "Who created KORA?"}
-  ]
-}
-```
+* DevOS
+* VUX
+* Kontyra Jobs
+* Tech Visionaries Network (TVN)
+* Kontyra Learn
+* Kontyra Projects
 
-### Endpoint
+## Author
 
-```http
-POST /v1/chat/completions
-Content-Type: application/json
-```
+**Heritage Oluwalonimi Oladoye**
+
+Founder of Kontyra
 
-Streaming responses are returned as SSE `data:` events compatible with OpenAI-style chunk consumption.
+Building technology that creates impact across Nigeria, Africa, and the world.

requirements.txt

@@ -9,5 +9,4 @@ torch==2.5.1+cpu
 accelerate==1.2.1
 sentencepiece==0.2.0
 python-dotenv==1.0.1
-peft==0.11.1
-
+peft>=0.12.0